CPU's in the future...

[taltamir]

Originally posted by: Greg04

Originally posted by: Hulk
Based on past progress where do you think we'll be as far as desktop processor performance in 10 years, 50 years, and 100 years? How far will the performance be pushed?

Will desktop processors be capable of "supercomputer" performance in the near future?

10 years ago the fastest computer of the day was the Pentium II 450. In CPUmark99 that processor did about 34. A fast (say 3.2GHz) quad core C2D running four instances of CPUmark99 could score about 1800. That's about a 50x increase in performance in 10 years. Could we see another 50x increase in performance in the next 10 years? 1M Superpi times in a fraction of a second? HD video compressed at 50x realtime? Compress a two hour movie in two minutes?

Of course these are all really rough estimates but I'm curious as to what other people in this forum think will happen in the next 10, 50, or 100 years? I can't even begin to estimate how fast cpu's will be in 50 or 100 years.

It doesn't matter how fast CPUs get, Microsoft will be there to trip up even the fastest one. I can see it now, WINDOWS AEON (WINDOWS 22) Minimum System Requirements: 100 Core AI Gel Cubes, 40PB Holographic Storage, and at least a 500TB RAM Array (with 1KW of power dedicated to lighting the Bill Gates laser memorial on the moon). Of course, consumers will continue to revolt against this bloatware, and XP will have reached Service Pack 279 (where *almost* all the bugs have been worked out).

Funny you should say that.. I have been experimenting with linux and solaris recently.. Linux is significantly slower then vista. And solaris is the slowest behemoth I have EVER seen. Solaris litterally chugs along at 1/4th the speed vista does. And I was wishing vista was as responsive as XP.. HAH!

 

[Hulk]

I am one of those people that is perhaps a little optimistic when it comes to thinking that OS bloat is coming down from one OS release to the next. Now I am NOT saying that Vista is leaner than XP, or XP leaner than 2000. What I am saying is that it seems that the increase in bloat from one OS to the next is decreasing.

Add that to the fact that the initial Windows releases were kind of overpowering the available hardware of the day. Of course this is just my take on the situation.

My first PC was a 486SX20 running Windows 3.1. It was slow just running the OS, opening Explorer, moving around files, etc... It wasn't until my P5-90 that Win3.1 felt kind of responsive. Then I moved to Win95 with the P5-90 and again Windows was slow. Then I upgraded hardware to a Celeron 450a (o/c) and Windows felt snappy again. Then Win98, which was pretty close to 95 but my cpu was now a PIII850 with faster everything in the box.

I stayed with Win98 until XP in 2002. XP on a PIII850 was not a good experience but it wasn't all that bad either. I'm STILL running WinXP with a C2D at 3.2GHz and XP is manhandled into shape by the C2D no problem.

My point is that cpu's and associated hardware seem to be outpacing the OS bloat. Again I'm being optimistic and I realize that. And of course it's not just OS bloat but bloat in lots of other applications.

One thing I wonder is how great all the this future cpu power will be for medical research and other scientific research.

 

[taltamir]

I agree with you Hulk... take vista and XP...
I remember when XP came out I could only reasonably afford 512MB of ram.. LATER I upgraded to 1GB and eventually the "massive" 2GB... Now with vista you need more ram. But I will say that vista 4GB is equivalent to 2GB XP, and 2GB vista to 1GB XP... and buying 4GB when vista came out was a lot cheaper then buying 2GB when XP came out...

The biggest issue I hear when it comes to vista bloat is that it copies files really slowly... this has nothing to do with bloat. MS rewrote their file copying algorithm to make it MUCH faster on network transfers under certain conditions, at the cost of making it MUCH slower on normal drive to drive transfers. That didn't sit well with people as 99.99999999% of people will not benefit from it. Stupid move.
They have been improving it though. But there is no option to revert it.

 

[lopri]

Originally posted by: Idontcare
At some point in the very near future this whole "x86 licensing embargo" should become a moot issue as it becomes public domain.

Can't wait

 

[Nathelion]

Originally posted by: lopri
What do you guys think of Intel's strategy? Obviously they realized their limit long time ago and have been pushing multi to many cores for some time. Now it's clear that they want 'x86 everywhere' for the company's future. (Atom, Larrabee, etc.) That agenda seems to be 'approved' so far because of Intel's superior manufacturing.

It is my understanding that as process nodes get smaller, the "cost" of the x86 decoder goes down in relation to the rest of the processor. While RISC vs CISC was a HUGE issue in the 80s and 90s, it's not anymore, because the footprint of the x86 decoder in terms of transistor count is constant or only slowly increasing, whereas the total transistor count of chips in general increases exponentially.
It follows that as process nodes shrink more, x86 becomes feasible in smaller and smaller devices, until it eventually overtakes all other ISAs (or, if something very unlikely happens, a different ISA could take over - but there will be a universal ISA, either way). So from that perspective, I think Intel's got the right idea. Unifying ISAs over more and more fields is great for Intel, since new competitors in the x86 space is an impossibility (x86 licensing issues), and they have their current competition well in hand.

Edit: The x86 IP protection is an interesting question, it would be nice if someone "in the know" could chime in. Although, even if patents expired on the older parts of the ISA, the newer parts might still be protected. Could you really build an x86 processor without SSE and still be competitive?

 

[toadeater]

Originally posted by: Modelworks
Organic has tons of potential.
Imagine 'feeding' your pc

It seems my PC has vomited.

Wait... no, it was the cat.

 

[taltamir]

Originally posted by: Nathelion

Originally posted by: lopri
What do you guys think of Intel's strategy? Obviously they realized their limit long time ago and have been pushing multi to many cores for some time. Now it's clear that they want 'x86 everywhere' for the company's future. (Atom, Larrabee, etc.) That agenda seems to be 'approved' so far because of Intel's superior manufacturing.

It is my understanding that as process nodes get smaller, the "cost" of the x86 decoder goes down in relation to the rest of the processor. While RISC vs CISC was a HUGE issue in the 80s and 90s, it's not anymore, because the footprint of the x86 decoder in terms of transistor count is constant or only slowly increasing, whereas the total transistor count of chips in general increases exponentially.
It follows that as process nodes shrink more, x86 becomes feasible in smaller and smaller devices, until it eventually overtakes all other ISAs (or, if something very unlikely happens, a different ISA could take over - but there will be a universal ISA, either way). So from that perspective, I think Intel's got the right idea. Unifying ISAs over more and more fields is great for Intel, since new competitors in the x86 space is an impossibility (x86 licensing issues), and they have their current competition well in hand.

Edit: The x86 IP protection is an interesting question, it would be nice if someone "in the know" could chime in. Although, even if patents expired on the older parts of the ISA, the newer parts might still be protected. Could you really build an x86 processor without SSE and still be competitive?

Well, the x86 is more then a single patent. It is protected by many different ones. Even if the core technologies go public domain, the improvements that everyone expect (SSE, etc) would go on forever.
Also, those things last almost forever.

I am shocked no country has revoked the patents and captured the technology in the interest of national security. It is a very VERY big issue to have all the computers in your country, the very foundation of your economy AND culture, AND communications, AND scientific research AND all systems of government be held hostage by a foreign corporation and be purchased from them in a black box format.
And if it is opened up, many different organizations could join in and accelerate the development rate.

But then again, you have people in the US congress who say things like "we should dip in acid the computers of everyone who has an mp3". So you can hardly expect the old and ignorant to grasp the importance of this technology, but the day is coming quickly where people who truly understand the computer world get in power.

 

[Idontcare]

Originally posted by: taltamir
I am shocked no country has revoked the patents and captured the technology in the interest of national security. It is a very VERY big issue to have all the computers in your country, the very foundation of your economy AND culture, AND communications, AND scientific research AND all systems of government be held hostage by a foreign corporation and be purchased from them in a black box format.
And if it is opened up, many different organizations could join in and accelerate the development rate.

If it is covered by patent then it is unlikely to be covered in all countries, but you can't make product and sell it in countries in which the product violates the patent protection in that country.

I've got some 15 patents thru my pervious employer (Texas Intruments) relating to process tech. The patent attorneys would always file first in US, then Japan and Taiwan. Then they'd sit on it for a while before launching a semi-global patent effort across Europe and the Asian countries. But we never filed in Africa. So technically someone could "steal" the IP and use it to make products that were only sold in those countries where we didn't have patent protection.

So obviously you have to be selective in your patent protection efforts and make sure you cover your relevant market spaces but not go to the extreme of investing $3M in attorney fees to protect every single patent in all 170+ countries.

(there are such things as "global" patents, but they were never favored by TI's lawyers for some reason)

 

[Lonyo]

Well, 50x performance in 10 years would be slower than current estimated pace (estimated by me)
2000: 1GHz P4
2008: 3GHz 8 Core Nehalem.

That's a 24x increase comparing clock to clock. Now lets say Nehalem can do 2x as much work per clock (something I believe is probably the minimum increase), and you have a 48x increase in theoretical power. In ~8 years. And that's pretty much a minimum. If we hit 16 cores before 2010 then that's a 100x increase in 10 years.

In 10 years from now, the minimum increase I would expect, assuming we continue with current technology, is 100x.
Of course, there will likely be some major breakthrough between now and then (like quantum or bio), although accounting for what this will mean is pretty much impossible, so I won't even hazard a guess.
And because of that, even thinking about 20 years and 50 years from now is not even worth it.

 

[BladeVenom]

In ten years we will have 32x more powerful CPU's. In 50 years we will have blown up most of the planet and any survivors will be counting with sticks. In 100 years we will have re-invented the abacus.

 

[nyker96]

I think x86 will come to an end in 50 years time. It's old and will eventually constrain the CPU development and newer instruction set standard will be eventually deviced and be practical enough to be implemented and will many new tech will see x86 as a bottleneck and eventually abandon it. With this said it may in fact shift the balance of power in CPU industry toward some new firm. Who knows.

 

[v8envy]

Originally posted by: nyker96
I think x86 will come to an end in 50 years time. It's old and will eventually constrain the CPU development and newer instruction set standard will be eventually deviced and be practical enough to be implemented and will many new tech will see x86 as a bottleneck and eventually abandon it. With this said it may in fact shift the balance of power in CPU industry toward some new firm. Who knows.

Except that Intel is not leaving the x86 instruction set static. They never have! I mean, how many people still program 'real mode' 8088 applications? Approximately 0.

Intel keeps extending the instruction set. 286, 386, 486, Pentium, 686, SSE, SSE2 ... SSE 4.3, AMD64. They add registers, enhanced addressing modes and other impressive chunks regularly. Yes, it may be possible for someone to make a clone of the 8088 royalty free (if it isn't already), but so what?

x86 really does mean the extended x86 instruction set code museum, complete with warts and undocumented implementation details.

 

[taltamir]

v8envy is right about it. x86 is pretty much an umbrela term for "every instruction set intel made in the last 20 years".
They will probably still call it that in 50 years when it uses completely different methods of processing data.

 

[myocardia]

Originally posted by: BladeVenom
In ten years we will have 32x more powerful CPU's. In 50 years we will have blown up most of the planet and any survivors will be counting with sticks. In 100 years we will have re-invented the abacus.

Haha, I thought I was the only one who thought that. Welcome to the club, BladeVenom.

 

[Idontcare]

Originally posted by: myocardia

Originally posted by: BladeVenom
In ten years we will have 32x more powerful CPU's. In 50 years we will have blown up most of the planet and any survivors will be counting with sticks. In 100 years we will have re-invented the abacus.

Haha, I thought I was the only one who thought that. Welcome to the club, BladeVenom.

It's a tad optimistic on how quickly civilization will recover the invention of the abacus though. We'll be doing good to have re-learned how to farm (agriculture) without the pesticides and benefits of fossil-fuel powered irrigation systems.

You'd think the Amish might do allright for themselves, except for the hordes of famished mobs that will stream out of Philadelphia and proceed to strip-mine the countryside for food during that first winter.

Planet of the Apes wasn't so farfetched when you think about it...

 

[taltamir]

Idontcare, you assume that somehow no educated person would survive and pass on their knowledge. It makes no sense for only CHILDREN to survive, adults are much better at survival. And there are plenty of young adults (20s) today who are sufficiently knowledgeable.

As for being the only one thinking it... Einstein said "I know not what weapons WW3 would be fought with, but I know that WW4 would be fought with sticks and stones" he was joking about WW4 ofcourse, since if you have only sticks and stones you can't have a WORLD war... But the point is, people have figured out humans are facing immanent destruction at their own hands many years ago.

 

[Idontcare]

Originally posted by: taltamir
Idontcare, you assume that somehow no educated person would survive and pass on their knowledge. It makes no sense for only CHILDREN to survive, adults are much better at survival. And there are plenty of young adults (20s) today who are sufficiently knowledgeable.

If someone, anyone, survives with knowledge intact regarding the abacus then the argument that humankind "rediscovers" something afterwards is invalid...you don't rediscover something that you are already knowledgable of.

I didn't make the statement we'd rediscover the abacus in 100 years. But in order for us to rediscover it, or anything else such as effective farming techniques in a stone-age world, then it presumes everyone with pre-existing knowledge of it had already perished.

 

[bryanW1995]

I think that we'll all be lighting candles and singing kum-ba-yah forever...

 

[Hulk]

Skynet.

 

[Viditor]

Originally posted by: bryanW1995
I think that we'll all be lighting candles and singing kum-ba-yah forever...

Do you think they will still have candles then?

 

[BadRobot]

I'm hoping that the Atom processor is the beginning of the end for craptacularly slow hand held devices (phones, PDA's, anything made by symbol, etc)

Faster hardware + intel's instruction sets better revolutionize hand held computing. If not I blame the programmers or manufacturers of devices for resisting change. Money grubbing bastards

'Small' computing is the future.

 

[myocardia]

Originally posted by: Viditor

Originally posted by: bryanW1995
I think that we'll all be lighting candles and singing kum-ba-yah forever...

Do you think they will still have candles then?

Viditor, haven't you ever been to a mall? I'm thinking the candles in a single Hallmark store would supply a fairly decent sized city for quite a few months.

 

[johnnyjohnson]

There's a physical limit on how small they make these integrated circuits. They will have to make chips bigger or go multi-processor. The days of single processor machines will seem quaint.

 

[Idontcare]

Originally posted by: johnnyjohnson
There's a physical limit on how small they make these integrated circuits. They will have to make chips bigger or go multi-processor. The days of single processor machines will seem quaint.

It's already quaint. How many single-core CPU's do you think Intel and AMD sell nowadays?

 

[james1701]

Originally posted by: taltamir
Idontcare, you assume that somehow no educated person would survive and pass on their knowledge.

Just because they have the knowledge does not mean they will have the capability to use it. If someone knows how to build an intigrated circuit, will they have the capability to build a foundry to make the parts to build it? If there is a mass disaster, it will be too hard to get all the people together with all the knowledge to get everything working again. I think everyone should start learning how to farm again just to be able to eat.